I was at a dinner recently with a group of senior leaders. The conversation moved, as it always does, to what was keeping people up at night. Within ten minutes, three of them mentioned restructures either underway or about to be announced. I asked each of them the same question: what’s the problem you’re trying to solve? The answers were vague. The timelines weren’t.

That exchange stayed with me, because in more than twenty years of leading and advising through organisational change, I’ve learned that a restructure is often the answer to a question nobody has properly asked. Sometimes the problem is structural, but often it isn’t. And by the time an organisation has committed to redrawing the org chart, it can be very hard to admit that the real issue lives somewhere else entirely.

Right now, many of the restructures I’m hearing about have a common driver: cut headcount to fund AI. That’s not a transformation strategy, that’s a cost exercise wearing transformation’s clothes. And it carries a particular risk, because the speed at which organisations feel pressured to move on AI means the human work that determines whether a restructure succeeds or fails is being skipped entirely.

The restructures that work

I’ve been inside many restructures that genuinely transformed organisations, not just rearranged them. The ones that worked shared a few things in common, and none of them started with the org chart.

They started with the customer and worked backwards. They got clear on the problem before they committed to the solution. They engaged people at all levels to pressure test the thinking and surface the problem areas that don’t show up in a consultant’s diagnostic. They communicated relentlessly, across every channel, not just the Town Hall. And critically, the leadership team got aligned upfront on the behaviours they themselves needed to role model for the process to succeed, not just as a project, but as a human experience.

What actually happens inside most restructures

Here’s what I’ve seen more often. Leaders who don’t believe in the restructure but have to execute it anyway. They’ve been told to get on the bus or get off it, so they get on the bus. They’re forced to choose between people who are valued, highly capable, and in many cases friends. They don’t believe the new structure solves the problem, but they are expected to be good corporate citizens, so they deliver the message and carry the personal cost quietly.

This is not anecdotal. Gartner reports that 75% of CHROs say their managers are more overwhelmed than ever, and one in four managers would opt out of their role entirely if given the chance. AI-driven productivity gains are coming with hidden oversight costs, and the people absorbing those costs are the leaders in the middle, the ones being asked to execute restructures they didn’t design, at a pace they didn’t set, while holding their teams together. The people being cut are frequently the ones carrying deep organisational knowledge, the kind of institutional memory and relational intelligence that no technology can replicate and that the organisation will need in order to make AI adoption actually work.

This invariably creates a low-level undercurrent of fear beneath the entire process. Not fear of incompetence, but something more subtle: the fear that if we slow down and are honest about what we’re really seeing, we’ll be accused of not moving fast enough. So everyone keeps moving. The restructure gets delivered. The milestones are met. And twelve months later, the organisation has fewer people, an AI investment it doesn’t yet know how to operationalise, and a culture that trusts leadership less than it did before.

The moment that changes the room

I’ve been in organisation-wide change processes where the leader has been handed a beautifully crafted approved script, looked at it, and torn it up. They then spoke from the heart instead. They treated people like trusted adults who deserved the truth so they could make their own decisions. Trust skyrocketed after that. Not because the message was easy, but because it was real.

That moment doesn’t happen by accident. It happens when leaders feel safe enough to be honest, and when the organisation has created the conditions for that honesty rather than penalising it. The restructures that actually change something are the ones where someone, usually a CEO or a Board chair with real conviction, creates space for a different kind of conversation.

Slowing down without losing momentum

The competitive pressure around AI is genuine. Nobody wants to be the company that hesitated while the market moved. But speed and thoughtfulness are not opposites. The organisations that will get AI right are the ones that slow down long enough to understand what they’re automating and why, which roles require human judgment that AI can augment but not replace, and where the institutional knowledge lives that they cannot afford to lose.

Practically, that means piloting AI in specific functions and learning what it actually changes before redesigning the whole organisation around an assumption. It means being honest with people about what is known and what isn’t. It means resourcing the People function to hold the human side of this properly, because a good HR team will manage a restructure in a way that keeps things moving with professionalism and empathy. People functions have been hollowed out in many organisations, and the work is increasingly outsourced to consultants, which can deliver structural outcomes but often leaves scar tissue of its own.

What Boards can do

Boards that govern this well ask a different set of questions. Not just whether the restructure will deliver cost savings, but what is being asked of leaders personally through this process. They ask what organisational knowledge is at risk of walking out the door. And in the context of AI, they ask whether the restructure is genuinely redesigning for a new way of working, or simply cutting headcount and calling it transformation.

The regulatory environment around psychosocial safety makes these questions more than good governance. They are an obligation. A restructure that creates prolonged uncertainty, erodes trust, or places unsustainable pressure on leaders and teams is a psychosocial risk. It deserves the same governance rigour we apply to financial risk.

The restructure that actually changes something

The structure is the architecture, but the leadership is what lives inside it. If we only redesign the architecture without asking leaders to examine what they are carrying into the new model, we will keep rebuilding the same house.

The organisations that get this right use the moment to have the conversations that should have been happening all along: about what the organisation actually needs from its leaders now, about what needs to be released, and about what kind of culture the new structure is being built to hold. With AI reshaping what organisations look like and how they work, those conversations have never mattered more.

That work is harder and slower than redrawing reporting lines. It is also the only work that makes the new lines mean anything.

If you’re leading a restructure right now:

If one in four of your managers would opt out of their role tomorrow, what does that tell you about the environment you’re asking them to lead a restructure inside? And if your restructure succeeded structurally but failed humanly, would you know?

If you’re governing one:

Are you confident this restructure is redesigning for a new way of working, or is it cutting headcount and calling it transformation? And is the person holding the human side of this resourced to do it properly?